Publications

Published as K.D.S.A. Chandrasiri or S. Chandrasiri. Full list on Google Scholar.

2027

CS REVIEW

Energy-Efficient Multi-Objective Cloud Workflow Scheduling: Taxonomy and Techniques

K.D.S.A. Chandrasiri, D. Meedeniya
Computer Science Review, 63, 101027, 2027.

Reviews energy-aware, multi-objective cloud workflow scheduling across heuristics, meta-heuristics and machine-learning methods, with an emphasis on Pareto awareness and Pareto-based evaluation metrics. It gives a unified taxonomy of workloads and constraints, compares simulation frameworks and datasets, and outlines open problems such as carbon-aware scheduling and explainable deep RL.

DOI

2026

IEEE ACCESS

Multi-Objective Cloud Workflow Scheduling with Two-Stage Deep Reinforcement Learning and Graph Neural Networks

K.D.S.A. Chandrasiri, D. Meedeniya
IEEE Access, 14, 34802–34829, 2026.

Proposes 2SD-GAT, a scheduler that encodes the workflow DAG with a graph attention network and makes each decision in two stages (pick a task, then a VM), trained with PPO on a preference-weighted reward over makespan, energy and QoS latency. On real-world Google cluster traces under dynamic conditions it achieves a 26.8% hypervolume gain and 4.5× lower IGD than the baselines, with low decision latency.

PaperDOICode
MSC THESIS

Energy-Efficient Cloud Task Scheduling Using Deep Learning

K.D.S.A. Chandrasiri. Supervisor: D. Meedeniya
MSc by Research thesis, University of Moratuwa, 2026.

Brings together the MSc research on learning-based, energy-aware scheduling of cloud workflows, from the DRL + GNN scheduler for dynamic workflows to the two-stage 2SD-GAT scheduler.

2025

SENSORS

Energy-Efficient Dynamic Workflow Scheduling in Cloud Environments Using Deep Learning

K.D.S.A. Chandrasiri, D. Meedeniya
Sensors, 25(5), 1428, 2025.

Combines a graph neural network with deep reinforcement learning (PPO) to schedule dynamic cloud workflows for both makespan and energy. In a CloudSim-based simulation it cuts makespan by up to 13.9% against HEFT, Min–Min and Max–Min while keeping energy use competitive.

PaperDOICode

2022

COMP. BIOL. CHEM.

CH-Bin: A Convex Hull Based Approach for Binning Metagenomic Contigs

S. Chandrasiri, T. Perera, A. Dilhara, I. Perera, V. Mallawaarachchi
Computational Biology and Chemistry, 100, 107734, 2022. Presented at the 20th Asia Pacific Bioinformatics Conference (APBC 2022).

Bins metagenomic contigs by clustering with convex-hull distance instead of Euclidean or L1 distance, using multiple large k-mers as high-dimensional composition features.

PaperDOICode